🎵 ELYSY | Eliz's Lyrics Synchronizer
AI-powered word-by-word lyrics synchronization — free, open source (MIT) and 100% local.
Everything you need for karaoke-perfect lyrics
Upload a song, paste or import its lyrics, and let AI do the timing — then fine-tune every word by hand.
AI Auto-Align
Demucs isolates the vocals, WhisperX produces word-level timestamps, and a matcher maps them onto your exact lyrics — preserving every line and word.
Manual Tap + AI
Tap Space at each line boundary, review and nudge the timing, then let AI refine word timing inside your tapped windows. Ideal for distortion or unusual phrasing.
Word precision controls
Hover any word for « < 00:23.45 > » buttons that nudge by ±0.1 s / ±0.3 s — with audio preview at that exact spot.
Undo / Redo
50-step history with Ctrl+Z, Ctrl+Shift+Z, Ctrl+Y and on-screen buttons. Also multi-line select with batch shift.
Projects
Save, Save As New and Load — self-contained project folders (audio + JSON) that you can move or back up freely.
Import LRC
Load existing .lrc files (standard or enhanced) to fix, re-time and upgrade them instead of starting from scratch.
5 export formats
Enhanced LRC, Enhanced SRT (word-by-word captions), Standard LRC, SRT and JSON — all with distinct filenames.
Local & private
Runs entirely on your machine. Your audio and lyrics never leave your computer. No accounts, no cloud, no tracking.
Two workflows, one perfect result
🤖 Mode 1 — AI Auto-Align (fastest)
Drop your audio
MP3, WAV, FLAC, OGG or M4A — up to 100 MB.
Paste or import your lyrics
One sentence per line — or click Import LRC to load an existing file.
Click Auto-Align
Wait ~30–90 s on a GPU (a few minutes on CPU).
Review
Play the song — words highlight as they are sung. Hover any word to adjust its timing.
Export
Enhanced LRC, Enhanced SRT, Standard LRC, SRT or JSON.
👆 Mode 2 — Manual Tap + AI (most precise)
Upload audio + paste lyrics
Then switch to Manual Tap mode.
Start tap recording
Press Space at the start of each line as the song plays.
Review the line timing
Nudge any line with « < > » or clear a mistap before continuing.
Refine words with AI
Word timing fills inside your tapped line windows.
Fine-tune & export
Adjust individual words, then export the format you need.
The pipeline under the hood
Four stages turn raw audio + lyrics into perfectly synced files:
Demucs — Vocal separation
AI source-separation isolates the singing voice from the instruments for far better accuracy.
WhisperX — Forced alignment
Whisper large-v3 transcribes; wav2vec2 aligns to phonemes for word-level timestamps.
Matcher — Lyrics mapping
Sequence alignment maps detected words onto your known lyrics, preserving every original line and interpolating anything missed.
Generator — Export
Produces Enhanced LRC, Enhanced SRT, Standard LRC, SRT or JSON.
System requirements
| Component | Minimum | Recommended |
|---|---|---|
| RAM | 8 GB | 16+ GB |
| GPU | None (CPU works) | NVIDIA 8+ GB VRAM (RTX 30/40/50) |
| Disk | 8 GB free | 12 GB free |
| Python | 3.10+ | 3.11 or 3.12 |
| OS | Windows 10/11, macOS, Linux | Windows 11 |
Move fast without touching the mouse
| Key | Action |
|---|---|
| Space | Play / Pause (or tap in manual mode) |
| ← → | Seek back / forward 3 s |
| ↑ ↓ | Seek back / forward 1 s |
| Enter | Run Auto-Align |
| Esc | Cancel tap mode |
| Ctrl+Z | Undo |
| Ctrl+Shift+Z / Ctrl+Y | Redo |
| Ctrl/Cmd + click (line) | Multi-select lines for batch shift |
| ⏮ | Seek to start |
Common issues, solved
| Problem | Solution |
|---|---|
| "CUDA not available" | Update NVIDIA drivers; re-run install.bat (CUDA 12.8 build). |
| Demucs slow / fails | Disable "Vocal separation" — alignment still works on the full mix. |
| Words missing or dotted | They are estimated — hover and nudge, or use Manual Tap mode. |
| Repeated chorus mis-timed | Auto-fixed by the monotonic pass; if not, nudge those lines manually. |
| Edits don't appear in export | Edits sync automatically; if unsure, Save then re-export. |
| Server won't start | Port 8765 busy? Run python backend\app.py --port 8777. |
| Install file-lock errors | Close the server + VS Code, pause Dropbox/antivirus, re-run install. |
ELYSY · MIT Licensed · Built with ❤️ for musicians, named after Eliz 🎤